Took a while since my last blog post but this one about capacitor’s needed some preparation time.

The capacitor story

In many designs (even some app notes) there are huge electrolytic or polymer cap’s to filter noise. And in many cases exactly those huge capacitors fail completely. They filter out the fundamental ripple of the switched mode power supply but when it comes to higher frequencies it seems they are not there anymore. But why?

Is my capacitor always a capacitor?

This is the most important and fundamental question. If the capacitor is capable of filtering out the fundamental waveform but not the higher harmonics then this leads to the assumption that the electrolytic capacitor I use is not always a capacitor.

For this reason I made a test setup to do a kind of proof of concept measurement for this assumption. Its a very simple test basically. With a resistor (150k) in series with a polymer-cap (15µF and 0,1Rser) I have built a first order low pass filter. For testing I set my frequency generator in my Keysight Scope to an offset of 3V and an amplitude of 2V (to keep the voltage across the capacitor always positive). Then I connected one channel to the input of the filter and one channel to the output of the filter – both AC coupled. For keeping the measurement as good as possible I generated a script to automatically step through the frequency range from 1Hz to 20MHz – which is the maximum I can get out of the frequency generator. Such a script should eliminate a lots of errors out of measuring by hand.

Frequency response of the low pass filter

Measurement analysis

The result of the measurement can be seen in the plot above. Theoretically the curve of the filter should start decreasing at 1 Hz already. I guess due to some measurement uncertainties this curve starts decreasing at about 10Hz. But more important is what happens after 10kHz. The filter settles after 10kHz (except this one dip which is cause by power supply noise of my printer which is nearby the test bench) at a maximum damping value – which is basically the value you get out of the voltage divider of the resistor and the series resistor of the cap. But at 1Mhz it starts to rise again – So my capacitor starts to act like an inductor. Such behaviour is quite normal within electrolytic and polymer capacitors and leads often to problems when it comes to EMC testing.

But how should I know in advance how this capacitor behaves? Simulators for the rescue! There are capacitor manufacturers like Kemet which provide a simulator which gives you an indication where you are going to land with your polymer capacitor but its just an indication because the wire length and the layout you have on your board are influencing this behaviour a lot!

What does that mean for power supplies?

Ok, what does this mean for your power supply design? As you might know a Buck converter is producing a high amount of noise at the input and a boost converter is producing a high amount of noise on the output. This noise is normally somewhere above the 20MHz I have shown here, so the ELCO and polymer caps will not be sufficient for damping such noise. In order to handle that I would recommend to put some smaller caps (ceramics) aside to this one the get rid of the noise.

When do I know i have enough ceramics for my noise? Well i would suggest a measurement with a LISN and a spectrum analyzer to determine which configuration is the best in your design. Without this it might lead to some over or underachievement in your design and either you have wasted board space or you have won some extra design cycles because your EMC noise levels are not where they should be.

Hope this analysis helps for your daily life in electronics!